A data-driven corrosion prediction model to support digitization of subsea operations

A data-driven corrosion prediction model to support digitization of subsea operations

Support vector regression

Principal component analysis

Artificial bee colony algorithm

Corrosion rate prediction

Subsea crude oil pipelines

Corrosion is an important factor leading to the failure of subsea process operations especially subsea crude oil pipelines. Developing a data-driven corrosion prediction model is urgently required by the digitization of subsea process system in the industry 4.0 environment, which is critical to improve the intelligent level of risk management of subsea process system. This paper proposed a new data-driven model based on hybrid techniques to model corrosion degradation of subsea operations. The model is built integrating three data-driven methods: principal component analysis (PCA), artificial bee colony algorithm (ABC) and support vector regression (SVR). The developed model is tested on the corrosion rate prediction of subsea crude oil pipelines. This model can realize effective prediction of corrosion rate. In the proposed hybrid model, PCA is used to reduce the dimension of corrosion influencing factors. The obtained principal components are selected as the input variables of the model. The ABC algorithm is adopted to optimize the hyper-parameters of the SVR. The model is trained using fraction of the historical data

subsequently, the model performance is tested on the remaining set of the data. A case study demonstrates the feasibility and effectiveness of the proposed model. The model is compared with the four different models SVR, PCA-SVR, PCA-GA-SVR, PCA-PSO-SVR. The PCA-ABC-SVR model performed superior in terms of prediction accuracy and robustness of results (MAE = 7.10 %

RMSE = 9.19 %

R2 = 0.976). The proposed model will serve as a useful online tool to support safety and digitization of process system.

413-421

153

Process Safety and Environmental Protection

2021

2021-09-01

Li, Xinhong

Zhang, Luyao

Khan, Faisal

Han, Ziyue

journalArticle

0957-5820

10.1016/j.psep.2021.07.031

https://www.sciencedirect.com/science/article/pii/S095758202100402X